Skip to content

Questions

Questions buyers ask. Before an AI project.

Straight answers to the questions we hear before a project starts: what we build, how we scope it, what it costs to run, who owns the result, and where people stay in control. Each answer links to the page with the details.

About Agents Autonomous

What does Agents Autonomous do?

We design, build, and hand over AI agents, workflow automation, and custom software for operations teams. Focused services cover reporting, document processing, customer support, and sales workflows, with evaluation and training available as separate scopes. Every engagement starts with one task and its owner.

Who will we work with?

Andrew Matusevich and Seva Rybakov, the co-founders, lead project conversations together and stay through the build. Andrew covers the business case, architecture, and operation; Seva covers agent engineering and integrations.

Where do you work?

We build for companies in the United States. Project conversations, reviews, and handoffs run online with your workflow owner involved from the first scope.

Do you publish client case studies?

Not yet. The examples on this site are illustrative or run on synthetic data, and they say so. Client references are shared in conversation when they exist and permission has been given.

Scoping a project

What makes a good first AI or automation project?

A repeated task with accessible examples, a clear output, and someone who can judge the result. Define its beginning and end, the tools involved, and the exceptions that need a person. That makes it possible to agree acceptance criteria before the build.

Do we need to choose between an agent, automation, or custom software first?

No. Describe the task and the tools it touches. Rules-based automation, a knowledge assistant, a tool-using agent, and a custom application solve different problems; the choice follows from how much judgment the task needs and how reversible its actions are.

Is this just ChatGPT for the team?

A licensed model on every desk can help one person. We look for the shared job, such as reporting, intake, follow-up, or review, where a better process helps the whole team and compounds.

What do we have to write down?

Enough that the next run knows the goal, the sources, the checks, and when to stop. That operating record is part of the handoff, not an extra slide deck.

How long will the project take?

The timeline is agreed for the specific scope. Access to systems, ready source material, unresolved business rules, and the availability of reviewers all affect delivery. The plan separates a feasibility prototype from a workflow ready for everyday operation.

Cost and ownership

How much does an AI or automation project cost?

Price depends on the workflow, source quality, integrations, permissions, and review requirements. A scoped proposal separates implementation from recurring model, hosting, platform, and support costs. Describe one task first so the estimate is based on the work and its dependencies.

What does an AI workflow cost to run?

Measure cost per accepted task: model calls, tools, infrastructure, retries, and review labor divided by the results a reviewer accepted. Token prices alone hide retries and review time. The guide and the calculator show the method.

Who owns and maintains the result?

Code ownership, licensing, hosting, access, and continuing support are terms to agree before work starts. The handoff names the owner for source updates, failures, credentials, and changes. Maintenance is part of the engagement only when it is explicitly included.

Which AI models and providers do you use?

We choose the model and provider for each project based on the task, the data-handling terms, and cost, and we name them in the scope. Different steps of one workflow may use different models.

Data, security, and control

Can we use the software and data we already have?

Assess the existing tools and sources first. APIs, exports, account permissions, data quality, and provider terms determine what can be connected. Begin with an outline or redacted examples; agree a sharing method before providing credentials or sensitive records.

Is our data used to train AI models?

That depends on the provider, the plan, and the configuration. Business and API terms from major providers generally exclude training on customer data by default, but the actual setting is checked and recorded for each service in the data path.

Do you replace the people who know the work?

No. They choose the job and they review what matters. AI prepares the middle. If a task should not be automated, we say so.

What can an agent do without asking?

Only what the permission matrix allows: typically reading approved sources and preparing drafts or records. Sending, filing, paying, deleting, and changing customer data wait for an approval gate unless the action is reversible and the owner has agreed otherwise.

Testing and results

How will we know whether the result is good enough?

Agree test cases and acceptance criteria with the person who owns the workflow. Check expected outputs, evidence, incomplete inputs, and permission boundaries. Review recorded failures as well as successful runs, then define the monitoring for the first rollout.

Which metrics do you use?

The ones that fit the task: correctness, completeness, harmful actions, review burden, latency, or cost per completed task. Thresholds are agreed for the project, not invented in advance.

Does passing tests mean the system cannot fail?

No. Tests cover the cases examined. Monitoring, human review, and a recovery plan are still needed for real operation.

How do we start?

Book a thirty-minute conversation and bring one workflow. If you want a head start, download the one-page project brief and answer its questions first.

A useful conversation starts here

Where does the work
get stuck?

If your question is not here, ask it on the call. Bring one workflow and we will answer for that case.

Discuss your project